NeuralFusion: Online Depth Fusion in Latent Space

TL;DR

NeuralFusion introduces an online depth map fusion method that learns to combine depth information in a latent space, effectively handling noise and outliers for improved 3D scene reconstruction.

Contribution

It proposes a novel learned feature-based fusion approach with a separate translator network, advancing beyond explicit scene representations like SDFs.

Findings

Outperforms state-of-the-art methods in noisy and outlier-rich scenarios

Handles high noise levels effectively in depth fusion

Demonstrates superior results on real and synthetic datasets

Abstract

We present a novel online depth map fusion approach that learns depth map aggregation in a latent feature space. While previous fusion methods use an explicit scene representation like signed distance functions (SDFs), we propose a learned feature representation for the fusion. The key idea is a separation between the scene representation used for the fusion and the output scene representation, via an additional translator network. Our neural network architecture consists of two main parts: a depth and feature fusion sub-network, which is followed by a translator sub-network to produce the final surface representation (e.g. TSDF) for visualization or other tasks. Our approach is an online process, handles high noise levels, and is particularly able to deal with gross outliers common for photometric stereo-based depth maps. Experiments on real and synthetic data demonstrate improved results compared to the state of the art, especially in challenging scenarios with large amounts of noise and outliers.